Abstract:With the development of automatic navigation technology, agricultural robots have been applied to all aspects of agricultural production. Agricultural robots can replace humans in activities such as spraying, fertilizing, and harvesting, reducing labor intensity and improving operational efficiency. Full coverage operation is one of the core contents of intelligent robot research, which involves many application fields such as agriculture, military, manufacturing, and civil. As a key technology in agricultural production operations, full coverage operation planning can help improve operation quality and resource utilization. However, in the full coverage operation, there are several challenges unresolved: obstacles identification is not accurate, hindering the working path of agricultural machinery; the area of the working area is omitted and the path is repeated, resulting in a waste of resources; the work efficiency of the single robot is low and it is unable to deal with complex full coverage problems. Starting with the problems existing in the full coverage operation planning, the construction of the environment model, robot path planning, and multi-robot cooperative task allocation was reviewed. Among them, accurate and reliable environmental map information helped to avoid static obstacles and improve operational reliability. Efficient optimization of path information helped to reduce missed areas and improve operational efficiency. The optimal task allocation scheme helped to reduce work time and waste of resources. Firstly, the environmental modeling methods were analyzed and compared with their limitations revealed, and optimization methods were put forward. Based on environmental modeling methods, the present situation of full coverage path planning algorithms at home and abroad was summarized, and the characteristics of related algorithms were pointed out. Then, the research progress of task assignment algorithms was discussed for multi-robot cooperative full coverage task allocation. Finally, the future development direction of the mobile robot full coverage task allocation was discussed. This research would help further improve the work efficiency and quality of the full coverage operation in agricultural production, and reduce the waste of resources. The research result provided an important basis for the realization of large-scale agricultural production in China.

Abstract:Aiming at the problems that the obstacles such as ditches, ridges, stones and electric poles in paddy fields make the rice transplanter unable to ensure the continuity of operation and the straightness of rice transplanting, a path planning strategy for rice transplanter around obstacles based on optimized artificial potential field method was designed. The relative distance between the realtime position of the transplanter and the target operation point was added as the judgment condition to dynamically change the size of the potential field. At the same time, a virtual local target point was set up to make up for the algorithm defects of the target point unreachable and local minimum point of the traditional artificial potential field method. The transplanter was simplified into a two wheeled vehicle model, the mathematical model of the steering system of the transplanter was established, and the expressions of the speed, heading angle and front wheel angle of the transplanter were obtained. The lateral deviation and heading deviation were used as the factors to judge the effect of path optimization. The steering controller used the compound fuzzy PID algorithm to control the rotation angle of the transplanter, continuously reduce the deviation between the ideal front wheel rotation angle and the actual rotation angle, and realize the optimization of the rotation angle. Using ultrasonic sensor to detect road obstacles in real time and RTK-GPS to update the position coordinates in real time, the steering control strategy of rice transplanter around obstacles was designed. The obstacle avoidance path control strategy of the optimized artificial potential field method was simulated by Matlab. The results showed that when the obstacles were not within the influence range, the maximum transverse position deviation of the straight-line tracking of the rice transplanter was 5cm, the average deviation was about 2cm, and the maximum obstacle avoidance transverse deviation was less than 0.5m. The optimized algorithm had good control accuracy and can avoid the problem of unreachable target points. Based on Yangma Vp6E rice transplanter as the experimental platform, the real vehicle experiment was carried out. The experimental results showed that when the rice transplanter ran at the speed of 0.5m/s, 1.0m/s and 1.5m/s, the maximum lateral deviation of the left side obstacle was no more than 1.2218m, the maximum heading deviation was 30.1491°, the average lateral deviation of the straight line tracking before and after the obstacle was 0.025m, and the average heading deviation was 3.12°. The maximum lateral deviation of the right side obstacle was not more than 1.2459m, the maximum heading deviation was 25.2294°, the average lateral deviation of straight-line tracking before and after the obstacle was 0.023m, and the average heading deviation was 3.36°. The designed obstacle avoidance method can meet the obstacle avoidance requirements of the transplanter during driving, and had good feasibility and robustness.

Abstract:In order to provide map and navigation service support for multi-agricultural machinery cooperative operation application scenarios, a multi-machine cooperative navigation service platform based on WebGIS was designed and developed. The platform consisted of two functional modules: GIS service and agricultural machinery scheduling. The GIS module provided web-side map services based on GeoServer and JavaWeb. While displaying the farm map and marking the location of agricultural machinery in real time, it also provided the visual display function of the navigation results of multiple agricultural machinery;the agricultural machinery scheduling module took the path planning algorithm and task allocation algorithm as the core, and responsible for providing navigation services. When the user provided the task list and invoked the service, it returned the task assignment and path planning results of each agricultural machine in GeoJSON format. In addition, in order to screen out the algorithm that met the platform requirements and had the best performance, algorithm performance comparison experiment were designed. The path planning algorithms based on A*, Bellman-Ford, Dijkstra, Floyd and SPFA were tested on three paths with short, medium and far navigation distances respectively, and the execution time and optimal path length were recorded and compared;for the task allocation algorithm, simulation comparison experiments under different task number scenarios were designed. The task allocation algorithms based on ant colony optimization and genetic algorithm were tested under the scenarios of 8, 10, 14 and 18 tasks respectively, and the execution speed and optimal path length of the algorithms were recorded and compared. The results showed that the path planning algorithm based on Dijkstra algorithm had the fastest execution speed under the premise of optimal results, and the average single execution time was 0.25ms. The task assignment algorithm based on genetic algorithm can effectively reduce the path cost of multi-machine collaboration. Compared with the randomly generated work sequence, the path cost was reduced by 50%~54%;compared with the algorithm based on ant colony optimization, the optimal path length was reduced by 19%~36%, the execution time was reduced by 51%~66%, and the average running time was within 1s. The developed multi-machine cooperative navigation service platform can basically meet the real-time requirements of multi-machine cooperative operation by using Dijkstra algorithm and genetic algorithm for path planning and task allocation.

Abstract:GNSS-based positioning and navigation has been widely used for agricultural robots in open unmanned farms. However, for the applications of semi-structured and semi-open agricultural scenarios, there may be temporary loss of GNSS received signals caused by occlusion of canopies in some areas, which will affect the positioning and navigation accuracy of robots and even harm crops or farmers. To solve this problem, a combined positioning method of GNSS and INS under the occlusion environment of agriculture was studied. The main work consisted of three parts: a mobile agricultural robot system was build up for the experiments of multi-sensor-based positioning and navigation, which consisted of hardware (track-layer mobile platform, GNSS receivers and INS, etc.) and software (ROS, remote control interface, etc.);an adaptive-coefficient Kalman filter based combined positioning algorithm was proposed. When the GNSS signal was unstable or denied, the new algorithm can switch to INS positioning adaptively based on Kalman filter, which carried out the optimal estimation for the robots’ location and gesture;experiments of the proposed combined positioning algorithm were conducted under practical scenes of agriculture, in which four different positioning methods (GNSS only, INS only, Kalman filter based combined positioning, and adaptive-coefficient Kalman filter based combined positioning) were compared to validate the effectiveness of the algorithm. Field experiments showed that in the process of combined positioning, compared with GNSS positioning, INS positioning and conventional Kalman filter fusion positioning, the positioning accuracy of adaptive-coefficient Kalman filter in the 30m×6m high shaded area of 100m×20m experimental area was improved by 62.1%,48.5% and 47.7%, respectively.

Abstract:According to agricultural machinery operation in headland pattern, through installing intelligent monitoring terminal in the agricultural machinery, data of spatial track of agricultural machinery was collected, the distribution frequency of azimuth angle was analyzed, the direction of agricultural machinery operation was calculated, the agricultural machinery operation lines were extracted, and model of row spacing of agricultural machinery operation was built. Furthermore, in order to verify the algorithm of row spacing of agricultural machinery operation, field experiments were conducted on June 25, 2021 at the National Precision Agriculture Demonstration Research Base in Beijing, China. The tractor installed automatic navigation system was selected, different row spacings were set by navigation system, such as 3.0m，4.0m and 5.0m. According to the set value of row spacing, the agricultural machinery operation experiment was carried out three times in headland pattern. The results showed that the average error was 3.07% between the set value of row spacing and model calculated value. The RMSE of validation was 0.14m. Overall, conclusion analysis of an example showed that its results were coincided with the active situation. Row spacing was an important parameter of farm machinery operation supervision, combined with the width of operation, it can provide data support to quality evaluation of agricultural machinery operation, such as overlapping and omission and so on.

Abstract:Aiming at the problems of unstructured and complex hilly orchard environment and low efficiency of conventional weeding methods, an orchard weeding robot chassis system was proposed.According to the hilly terrain and landform environment of the orchard, the vehicle body control mode and the overall structure scheme of the weeding robot chassis were determined, mainly including the hydraulic transmission system, and electrical control system, etc. With motion control as the core, in order to improve the control accuracy of the hydraulic system, the supporting electrical control system, remote control receiver and CAN communication protocol of the weeding vehicle were designed, which were composed of the main controller, remote control receiver, motor drive module and navigation module, and the software flow was determined. The motion controller was designed based on the active disturbance rejection control algorithm. The robot track speed was determined by the valve angle of the hydraulic system, it can be seen that the dynamic performance of the motor directly connected to the control valve was the motion control target. The Simulink simulation model was established by taking the motor angle of the hydraulic valve as the object and the speed, disturbance and output as the state variables. The simulation results showed that the ADRC reduced the adjustment time by 0.42s, the overshoot by 11.5% and the stability time by 0.14s compared with the PID control. Ground experiments were arranged to verify the effectiveness of the robot system. The experiments showed that the average walking speed of the weeding robot using active disturbance rejection control combined with navigation function was 6.2km/h, the mean square deviation was 0.037km/h, the operation efficiency was 0.51hm2/h, and the average effective weeding rate was 97.46%. It can walk normally on a 25° slope, and the standard deviation of tracking error for the navigation path was 4.732cm. The motion control response was timely, which can improve the safety and accuracy of weeding operation.

Abstract:Aiming at the problems of poor uniformity of low sowing quantity precision metering device of vibrating rice seedling tray at present, which is difficult to provide a single row of stable seed flow, a piezoelectric vibrating homogenizing device was designed. The structural parameters of each component were determined by analyzing the vibration principle of piezoelectric vibrator, the dynamics of vibrating plate and the steering of rice seeds. The structural parameters of the vibrating plate were optimized. Taking the depth of the seed storage box, the angle of the steering groove and the direction of vibration as the test factors, the optimization test was carried out in combination with the Box-Behnken test scheme. The results showed that the interaction of the angle of the steering groove, the depth of the seed storage box, the direction of vibration and the angle of the steering groove had a significant impact on the test results. When the depth of the seed storage box was 8mm, the angle of the steering groove was 49°, and the direction of vibration was 29°, and the coefficient of variation of seed evenness was 17.91%. The relationship between the input voltage and the amplitude was further determined by the acceleration measurement bench test. Under the optimal structural parameters, the variation coefficient of uniformity, the qualified rate of sowing and the missed rate of sowing were 18.20%, 94.65% and 0.67%, respectively. The seeding performance test at different uniform seeding speeds showed that when the working voltage was 130~180V, the qualified seeding rate was no less than 94.17% and the missed seeding rate was no more than 0.83%. The adaptability test results of different rice varieties showed that under the three working voltages of 130V, 150V and 170V, the qualified seeding rate was no less than 94.17%, and the missed seeding rate was less than or equal to 1.0%, which met the requirements of precision seedling raising and sowing of super hybrid rice.

Abstract:In order to get the optimal solution by comparative tests, the simulation analysis was widely used during the development process of air suction seed metering device. The gas-solid coupling simulation method was compatible because the single grain rate of air suction seed metering device was reached by different air pressures, and the process was filled with variable flow field and moving particles. The difference of contacting parameters between the simulation particle and the real particle would be bigger when the bonding particle model was implied during the coupling of Fluent. In order to solve the problem, the reasonable variable range during the process of calibrating the discrete element parameters was reached with the measured values of contact parameters between maize species. The steps were as follows: the rolling friction and the static friction of corn-corn were measured by slope method and energy conservation method of high speed photography. The linear function between the zero level of variable range and the measured value was established (the best calibration range of zero level was equal to the measured coefficient value of coefficients). In order to find the optimal coefficient value, six groups of different coefficients were selected to solve the variable range, and the response surface optimization method was applied to optimize and calibrate the corn interspecific contact parameters with the real measurement of corn accumulation angle as the objective quantity. The accumulation angle of simulation was obtained by calibrations parameters and compared with the real accumulation angle got by lifting tests. The relation between curve of accumulation angle was calculated by different coefficients values. The value of optimized parameter during the calibration range was the most accurate when the coefficient was 0.1977. The best combination of fixed parameters was as follows: the static friction coefficient between corns was 0.031, the rolling friction coefficient between corns was 0.0039. There was no significant difference between the simulation results of the simulation results of the best parameter with the real experiment. The results showed that the contact parameters of the calibrated discrete element of corn were reliable. Those parameters and calibration methods can be set as references for the selection of calibration parameter range in the simulation process of follow-up air suction seed metering device.

Abstract:The AquaCrop model driven by water was used to study the simulated yield of full film double ridge sowing technology of corn in Dingxi City from 2016 to 2020, and I1 (bare land planting), I2 (narrow film planting with 45% mulching rate), I3 (wide film planting with 81.8% mulching rate) and I4 (full film double ridge planting with 100% mulching rate) were established respectively. The yield superiority and environmental adaptability of the four models were compared and analyzed, The relationship between sowing date, rainfall and soil moisture was obtained. The results of AquaCrop simulation showed that the model was suitable for simulating dry farming in Dingxi City. The Pearson correlation coefficient (r) between the output simulation value and the measured value of I4 model was greater than 0.91, the root mean square error (RMSE) was 0.1~0.24, the normalized root mean square error (CV(RMSE)) was 1.66%~2.10%, the Nash efficiency coefficient (EF) was greater than 0.9, and the consistency index (d) was greater than 0.94. The average temperature of the best sowing date in Dingxi City was stable at about 15℃ (about April 15-25 every year). The yield after sowing was the highest in this period. The yield, aboveground biomass and water productivity of the planting mode I4 model were 84.01%, 19.79% and 101.13% higher than that of the planting mode I1 model, 82.26%, 19.74% and 85.47% higher than that of the planting mode I2 model, and 63.26%, 14.80% and 82.63% higher than that of the planting mode I3 model, respectively. The total soil water content of I4 model in dry years was more than 90% higher than that of I1, I2 and I3 models, and more than 80% higher in wet years. From 2000 to 2020, the water consumption was higher than the effective rainfall for seven years, and the water consumption was lower than the effective rainfall for 13 years. The simulation results showed that the water consumption of the full film double ridge sowing technology was equal to the rainfall, and the soil water would not be overdrawn.

Abstract:Aiming at the problems of high missing index and non-precision seeding due to the small theoretical seeding spacing, small qualified range and internal and external interference of the system during precision seeding of mung bean, an electric-driven mung bean precision seeder control system based on IGWO-LADRC was studied. The modeling of the seeding motor was proposed, and an improved grey wolf optimizer (IGWO) was proposed to tune the parameters of the linear active disturbance rejection control (LADRC). The advantages of intelligent control system were tested by comparing the four seeding motor control strategies of empirical tuning PID, empirical tuning LADRC, GWO-LADRC and IGWO-LADRC. Simulation experiment showed that the electric-driven mung bean precision seeding control system based on IGWO-LADRC algorithm had a fitness of 1.5320, no overshoot, a settling time of 0.57s, almost no steady-state error, a disturbance recovery time of 0.35s, and no oscillation phenomenon and no steady-state error after reached the steady state again. The results of the seeding performance experiment showed that compared with the traditional chain-driven seeding,qualified index was increased by 2.75 percentage points, multiple index was decreased by 0.46 percentage points, missing index was decreased by 2.22 percentage points, precision index was decreased by 8.91 percentage points, and the precision index of qualified seeding spacing was decreased by 7.89 percentage points,compared with the traditional PID electric-driven seeding, qualified index was increased by 2.20 percentage points, multiple index was decreased by 0.37 percentage points, missing index was decreased by 1.30 percentage points, precision index was decreased by 7.28 percentage points, and the precision index of qualified seeding spacing was decreased by 4.47 percentage points, all of which met the national standards.

Abstract:Naked seedling transplanting is the main planting method of sweet potato in China, which requires high mechanized transplanting. At present, the domestic sweet potato transplanting machinery is relatively lacking, the existing equipment has a low degree of automation, and most of them adopt manual seedling feeding, which leads to high labor intensity in the field and low quality of mechanized planting. Combined with the agronomic requirements of planting naked sweet potato seedling, an automatic transplanter for naked sweet potato seedlings was designed based on the pretreatment seedling belt feeding device and the flexible disc planting device, According to the requirement of automatic operation control of feeding device, flexible disc planting device and watering device of sweet potato bare seedling transplanter, a control system of sweet potato bare seedling automatic transplanter based on CAN bus was designed, which can complete rotary tillage, ridging, ditching, automatic and orderly seedling feeding, planting at fixed plant distance, pressing soil covering, automatic watering, ridge repairing and other operations at one time. Field experiments showed that when the distance of target plant spacing was 25cm and the working speed was 0.25m/s, 0.35m/s and 0.45m/s, the coefficient of variation of planting plant spacing and the qualified rate of planting depth all met the standard requirements, the variation coefficient of plant spacing and the qualified rate of planting posture were affected by the working speed greatly, but the planting depth was less affected by the working speed. When the working speed was 0.25m/s, the working performance index was better than 0.35m/s and 0.45m/s. The average variation coefficient of plant spacing was 10.16%, the average qualified rate of planting depth was 95.56%, and the average qualified rate of planting posture was 90%. The research result can provide a reference for the theoretical research and innovative design of mechanized and automatic transplanting machinery for naked sweet potato seedlings.

Abstract:For the plug-in seedling picking mechanism, when the seedling claws are inserted into the pot to pick up the seedlings, the success rate of seedling picking is low and the pot is broken due to the double influence of the adhesion between the pot and the plug and the poor packing. In order to solve the problem of high rate, a combined seedling picking technology with top clamping and pulling was proposed, the structure and working principle of the seedling picking device were expounded, and the experimental research on the combined top clamping and pulling seedling picking was carried out. Firstly, the 72-hole and 128-hole cucumber plug seedlings were used as the test objects, and the adhesion of the cucumber seedlings at different top seedling speeds (10mm/s, 20mm/s, 30mm/s and 40mm/s) was tested through the top pressure off-disk adhesion test. The test results showed that the top seedling speed had little effect on the adhesion force and the detachment displacement of the seedling pot, and the adhesion force was positively correlated with the detachment displacement of the seedling pot. The value was distributed between 5.5mm and 6.9mm. Considering the diameter of the seedling tray drain hole and the test results of the top pressure off tray adhesion test, it was determined that the diameter of the mandrel was 6mm, and the displacement of the top seedling of the mandrel must be greater than 5mm. Secondly, taking the 72-hole cucumber plug seedlings with a growth cycle of 25d as the test object, three modes of seedling collection were carried out, namely, the first top and then the top, the side top and the side, and the first plug and then the top. The seedling success rate and pot body integrity rate were the highest. Finally, taking the ejection displacement of the ejector rod, the depth of inserting the seedling claws into the seedling pot and the speed of inserting the seedlings as the test factors, a three-factor and three-level orthogonal test was carried out. The optimal working parameter combination was that the jacking displacement was 15mm, the depth of inserting the seedling pot was 35mm, and the speed of inserting the seedling was 225mm/s. The success rate of seedling removal under this combination was 94.12%, and the complete rate of the seedling pot was 94.12%, which satisfied the automatic seedling quality requirements.

Abstract:Oat is a kind of grain and forage crop, which is an important substitute for planting structure adjustment and an important variety to improve dietary structure in China. In view of the low level of mechanization of oat production, uneven development of various links, backward basic research and low degree of industrialization in China, which affect the development of oat industry. The current situation of oat industry in the world and China was expounded, and the weak links in the whole-process mechanization of oat production was analyzed from the perspective of the whole-process mechanization technology system. The development status of oat field breeding mechanization technology and equipment, mechanized tillage technology and equipment, mechanized sowing technology and equipment, mechanized field management technology and equipment, mechanized harvesting technology and equipment at home and abroad were summarized and analyzed. In view of the shortcomings in the development of the whole-process mechanization of oat production in China at the present stage, it was suggested to conduct in-depth research on basic research, diversified technology and equipment development, new technology promotion, standard formulation, industrial economic benefits, automation and intelligent equipment, so as to boost the mechanization process of oat industry and promote the standardization and industrialization of oat industry. It was expected to provide reference for the development of oat industry in China.

Abstract:Aiming at the problems of imperfect mechanized production system, complex and diverse production models, and lack of systematic evaluation in the hilly and mountainous areas of Southwest China, taking the main body of agricultural management as the research object, a “farmland + agricultural machinery + agronomy + information” was constructed from the perspectives of moderate scale, high production efficiency, and ecological friendliness. Four-integrated evaluation system of mechanized production mode in hilly and mountainous areas of Southwest China, which included six dimensions and 15 tertiary indicators, including suitable mechanization of farmland, moderate scale operation, quality of agricultural machinery equipment, degree of intelligence of agricultural machinery and equipment, agricultural machinery production efficiency and farmland health. The weights were determined by the AHP and CRITIC methods;four typical models of wheat/corn mechanized production in Southwest China were selected for evaluation and comparison based on farmland endowment, mechanical equipment as the core, and scale benefit as the guide. The results showed that under the mechanized production mode (M1) of the key links of ordinary farmers, the plots are small and the terrain is undulating, and the machines and tools can only use small and low-efficiency machinery;the application rate of chemical fertilizers and pesticides was higher than the standard value, and its operation scale and development mode cannot be used. The mechanized production mode (M2) of family farms with strip compound planting realized the mechanization of farmland contiguous operation and field production. However, the technology of strip compound planting and harvesting equipment was not yet mature, resulting in the overall operation efficiency of the model being lower than that of the cooperative model. The model can improve land utilization and increase grain production, and it was suitable for promotion in family farms;under the “full mechanization + digitalization” production mode (M3) of cooperatives, large and medium-sized machinery can give full play to operational efficiency and fuel efficiency in farmland after mechanization transformation. Efficiency advantage, at the same time, post-production drying and primary processing improved grain quality and efficiency, and realized the mechanization of the whole process of pre-production, production and post-production. In addition, the digital management system effectively improved the efficiency of agricultural management and equipment use. This model was suitable for cooperatives in hilly and mountainous areas. The whole-process mechanized production mode (M4) of large-scale cooperatives’ planting and breeding cycle realized an agricultural industrial chain combining planting and breeding on the basis of the whole-process high-efficiency mechanical empowerment. The silage straw produced by cooperatives was sold to dairy farms for production. The feed was processed, and the organic fertilizer produced by the dairy cows was supplied to the cooperative for recycling. This model realized ecological and economic farming and was suitable for promotion in some large farming counties. The comprehensive evaluation values of the four modes were 0.31, 0.67, 0.86 and 0.79, and the order from large to small was as follows: M3, M4, M2 and M1;the evaluation results were in line with the reality, and the index system can objectively evaluate the characteristics of each mechanized production mode, which can be used for the selection and improvement of the mechanized production mode of each operating entity in the hilly and mountainous areas of the southwest, which provided a theoretical basis, and it was further improved in practice.

Abstract:Due to the particularity of topography, farmland scale and land conditions in hilly and mountainous areas, there is a lack of scientific selection methods for wheat and corn harvesting machinery, especially the insufficient consideration of terrain conditions and safety, resulting in poor adaptability and passability of some harvesting machinery in hilly and mountainous areas and high loss rate. In order to improve the adaptability of wheat and corn harvesting machinery, reduce the loss rate of machinery, achieve safe, green and efficient production, and ensure the warehousing of grain particles, adhering to the basic principles of advanced application, light and efficient, green environmental protection, safety and comfort, based on the theory of agricultural machinery selection, through field research and expert consultation, Based on the combination of qualitative analysis with quantitative analysis, combining theory with practice, the integration of agricultural machinery and agronomy, the method of analytic hierarchy process (AHP) was used to construct the energy consumption, operation efficiency, operation effect and operation performance, comfort, security, and other six secondary index, totally 16 threelevel index of hilly wheat/corn harvest machine selection evaluation index system. According to the expert survey method and related research to determine the weight of the evaluation index. Based on field experiment and field investigation, six wheat harvesters, including J-2.5 (Ⅰ), L-3.5 (Ⅱ), W-4.0 (Ⅲ), S-4C (Ⅳ), A-5.0ZA (Ⅴ), W-6.0EA (Ⅵ), were evaluated and verified. The results showed that the comprehensive evaluation values of six wheat harvesters were 0.71, 0.43, 0.50, 0.73, 0.45 and 0.39, respectively, and the order of effectiveness was Ⅳ,Ⅰ,Ⅲ,Ⅴ,Ⅱ,Ⅵ. Evaluation results accorded with the actual use of farmers, showed that the evaluation index system of science and applicability in hilly and mountainous Southwest China was good, for the relevant management departments and farmers to provide a scientific basis for wheat harvest machine type selection, retrofit for farm machinery scientific research units, production enterprises, to create suitable for hilly wheat corn harvest machine to provide the reference.

Abstract:The current corn cleaning loss monitoring sensors are affected by various problems such as the variety of cleaning products and the complex environmental noise, and the monitoring accuracy is difficult to meet the actual needs. In order to solve this problem, a clearing loss monitoring sensor based on PVDF piezoelectric sensitive element was designed to separate the vibration, industrial noise and stray signals in the collected signal. A minimum energy criterion based on DSP electronic signal processing was proposed. The EMD denoising method used the decomposition order corresponding to the minimum energy point of the IMF component as the signal-to-noise boundary point. The amplitude discrimination circuit identified the impact signal and calculated the loss rate. In order to verify the feasibility of this method, the signal with Gaussian white noise was simulated for denoising. Compared with wavelet denoising, low-pass filtering and moving average, the Matlab simulation results showed that the EMD denoising method based on the minimum energy criterion had the smallest root mean square error (RMSE), the highest signal-to-noise ratio (SNR), and the processed signal was the closest to the original signal. Changing the signal-to-noise ratio of the original simulation signal further verified that the results obtained by this method were always optimal. In order to verify the accuracy of the method, the corn kernels and miscellaneous mixtures with loss rates of 0, 5%, 10%, 15% and 20% were used as impact samples. Compared with the experimental data obtained by three denoising methods, wavelet denoising, low-pass filtering and moving average method, the average error of the minimum energy criterion EMD denoising method was reduced by 2.12 percentage points, 4.40 percentage points and 6.52 percentage points, respectively. The research result was of great significance for improving the detection accuracy of corn cleaning loss rate, especially the research on denoising methods in the process of signal processing.

Abstract:Sweet potato vines grow lush, and crawl seriously, which are hard to be separated. Manual cutting of vines was characterized by high labor intensity, high cost and low efficiency, which affected the normal harvest operation. For the problems of transport device blockage, large power consumption and difficulty of collecting operating parameters in the harvest process,the sweet potato vines harvesting test bench was designed. The test bench consisted of feeding device, cutting table device and control system, and the feeding velocity, the clipping velocity and the cutting velocity can be adjusted. Sweet potato vines cleaning rate, cutting force, and cutting torque were the main indicators of evaluating the test bench.The tests were grouped the response surface tests and the verifying tests. The testing factors were the feeding velocity (ranging from 0.3m/s to 0.8m/s), the ratio of clipping velocity to feeding velocity (ranging from 1.03 to 1.87) and the cutting velocity (ranging from 0.96m/s to 1.64m/s), and five levels was set for each factor. The response surface testing scheme designed with the central composite design (CCD) method was a three-factor five-level testing scheme. The mathematical model of the response surface was established, and the influence of each factor on the bench operation performance was analyzed and optimized. The result showed that when the feeding velocity was small, the ratio of clipping velocity to feeding velocity was high and there was a moderate cutting velocity. As such, the overall impact trend of the vines cleaning rate was high. When the cutting velocity was high, the cutting force and the cutting torque was small. The feeding velocity and the ratio of clipping velocity to feeding velocity had no significant impact on the cutting force and the cutting torque. The experiment results also showed that the order of the factors affecting of the vines cleaning rate was ratio of clipping velocity, feeding velocity, and cutting velocity. The order of the main factors affecting of the cutting force and the cutting torque was cutting velocity, ratio of clipping velocity, and feeding velocity. It was concluded that the optimal combination of the working parameters of the bench was as follows: 0.55m/s for the feeding velocity, 1.48 for the ratio of clipping velocity to feeding velocity,1.50m/s for the cutting velocity, 91.0% for the cleaning rate,152.89N for the cutting force, 5.87N·m for the cutting torque. By comparing the mathematical model and the experimental result, the error was less than 5%, which meant that the model established was reliable and could be used for optimization.

Abstract:Aiming at the problems of low efficiency and high damage rate of artificial harvesting in domestic apple orchards, an apple harvesting platform conveying device was designed to realize the automatic conveying of apples. Firstly, through the kinematic analysis of apple conveying process, the main factors affecting apple mechanical damage and the test value range of each factor were determined. Secondly, according to the Box-Behnken test design method, taking the conveying speed, the dip angle of the fruit tray and the shape of the fruit tray as the test factors and the damage rate as the test index, the operating parameters of the conveying device of the apple harvesting platform were tested and studied, and the regression model between the test index and the test factors was established. Finally, the influence law of each factor on the test index was analyzed, and the test factors were comprehensively optimized according to the regression model. The results showed that when the conveying speed was 0.2m/s, the dip angle of the fruit tray was 30°, and the shape of the fruit tray was a circular platform, the apple damage rate was 3.69%, which was significantly lower than the mechanical damage of the apple conveying device without parameter optimization, and the conveying effect was the best. The results could provide a reference for the structural optimization of apple orchard harvesting platform and the control of transportation parameters.

Abstract:It is important to detect grain moisture content for grain yield monitor, grain accurate possession and precision agriculture. It limits the accuracy and stability of grain moisture on-line detection for the dynamic condition of combine harvester.A grain moisture on-line detection device on combine harvester was developed based on the principle of dielectric property. The dynamic constant volume sampling mechanism was proposed. With the help of the sampling mechanism, a method of dynamic continuous sampling and static interval measurement was also proposed. It could detect grain moisture on the operation condition of combine harvester. The device consisted of mechanical dynamic sampling part, motor control module, sensor module, data acquisition module, GPS module and a display terminal. Sensor module included moisture sensor, temperature sensor and grain height level sensor. In June 2020, experiments were carried out at Beijing Xiaotangshan National Demonstration Station of Precision Agriculture. Experimental results showed that under the static condition, grain moisture on-line detection error was within 3%. Grain moisture detection model was established integrated of permittivity and grain temperature. The dynamic experiment showed that the correlation coefficient between the predicted grain moisture value and that of standard value was 0.92. The relevant error was less than 5%. It could satisfy the practice need.

Abstract:The driving torque of the oscillating slat shovel harvesting device is determined by four groups of single torque coupling, and the maximum driving torque and torque fluctuation directly affect the power matching and effective power output of the tractor. Analytical equations and discrete element simulation were used to analyze the characteristics of the driving torque waveform of the harvesting device. The single torque showed a double peak per cycle and the strength of adjacent cycles. Due to the vibration balance between groups and soil viscoplasticity, the driving torque showed a double peak per cycle and a slight change in adjacent cycles. The peak value was about 1.2~2.3 times of the single torque. The maximum driving torque, torque fluctuation and specific power consumption were taken as the driving torque characteristics and influencing factors of the harvesting device. Based on the force analysis of the working monomer, the monomer torque, driving torque and specific power consumption were established as the indexes, and the four-factor and three-level Box-Behnken simulation experiment was carried out. The experimental results showed that the amplitude, vibration frequency, excavation depth, forward speed, and the interaction term between amplitude and vibration frequency had significant influence on the maximum driving torque. Under the fixed working condition, when the amplitude was 8~10mm and the vibration frequency was 8~9.67Hz, the maximum driving torque had a better value. The vibration frequency, excavation depth and their interaction were the main factors affecting the torque ripple. Under fixed conditions, when the vibration frequency was 8.67~10.67Hz, the torque ripple had a better value, and the excavation depth can be 350~450mm. The vibration frequency, amplitude, forward velocity, excavation depth, and the interaction term between vibration frequency and forward velocity were the most important factors affecting the specific power consumption. Under fixed conditions, when the forward velocity was 0.35~0.45m/s, and the vibration frequency was 7.67~9.67Hz, there was a numerical optimal value of specific power consumption. The field test results of liquorice harvester showed that when the excavation depth was 450~500mm (condition 1), the driving torque showed a double peak per cycle and the adjacent strength changed, the maximum driving torque was 797.17N·m, the torque fluctuation was 2.54, the specific power consumption was 122.06kJ/m3, and the liquorice harvest rate was 96.42%. With the excavation depth reduced to 350~400mm (condition 2), the strong and weak cycles of driving torque were not obvious. Compared with condition 1, the maximum driving torque was decreased by about 39.44%, and the torque fluctuation was decreased by about 27.95%, and the specific power consumption was basically the same. The oscillating slat shovel harvesting device had good torque balance ability, which can provide reference for the research on vibration drag reduction and energy saving harvesting of rhizome crops, especially deep rhizome crops harvesting equipment.

Abstract:Data mining technology based on cluster analysis can promote the precision production, fine management, and precise marketing of agriculture, which is of great value to realize the precision of agriculture and then promote the efficiency and modernization of agriculture. The connotation and methodological system of data mining technology based on cluster analysis were reviewed, including feature selection and feature extraction, distance metric, clustering algorithm classification, and clustering performance evaluation index;and then the current research on the application of cluster analysis in five major directions of agriculture—plant and animal genetic breeding data mining,precision management of farmland zoning, agricultural product quality evaluation, market segmentation of agricultural products, and farmer heterogeneity analysis were combed;finally, a summary of cluster analysis in agriculture was presented, and an outlook was given based on the actual needs in agriculture and the development of cluster analysis technology, which provided insight into the theoretical research and in-depth application of clustering technology in agriculture.

Abstract:In order to analyze the morphological structure of cherry tree leaves and provide a theoretical basis for the light distribution of cherry tree canopy and the shaping and pruning of cherry fruit trees,a method for reconstructing cherry tree leaves was proposed based on improved Harris corner detection and Nurbs curve. The obtained original cherry tree leaf images were preprocessed by median filtering method and classical edge detection algorithm to obtain leaf contours. An algorithm was proposed to reconstruct the contour, aiming at the problems that a closed surface cannot be formed when the contour was reconstructed by the traditional NUBRS curve principle, and the feature points interfere with each other. The algorithm first divided the feature points into left and right parts, the contours of the left and right sides were reconstructed respectively, and then the two were connected to obtain a complete contour. Secondly, the improved Harris corner detection method was used to extract corner points as feature points. Thirdly, by detecting the degree of similarity between the gray value of the center point of the window and the gray values of other pixels in the surrounding n neighborhood, the difference between the gray values was calculated to set a threshold, and the corner points were extracted according to the threshold range. Finally, a virtual leaf was constructed according to the virtual contour. Experimental analysis showed that the improved algorithm greatly reduced the amount of useless computation, and the average consumption time was reduced from 4.61s to 2.30s. Based on the improved algorithm, the edge shape of cherry tree leaves can be perfectly maintained, which provided technical support for the calculation of light distribution in the cherry tree canopy.

Abstract:The relationship between soil spectral reflectance and soil electrical conductivity was expressed indirectly by using the relationship between soil water content and NIR spectral soil reflectance and soil electrical conductivity with soil water content as an intermediate variable. There was an exponential relationship between soil water content and soil spectral reflectance, and a linear relationship between soil water content and soil electrical conductivity, and the relationship between soil spectral reflectance and soil electrical conductivity was obtained by eliminating the intermediate variable (soil water content). The exponential prediction model and the logarithmic prediction model were established and validated respectively by taking the soil moisture sensitive band 1450nm as the research object to study the prediction model of soil electrical conductivity. There were 72 samples in the experimental modeling set and 48 samples in the validation set, and the R2 of the logarithmic prediction model of soil electrical conductivity reached 0.80, and the R2 of the exponential prediction model of soil conductivity reached 0.85, both of which can satisfy the estimation of farmland conductivity, but the prediction effect of the logarithmic model was not satisfactory in the lower range of soil conductivity, so the prediction effect of the exponential prediction model of soil conductivity was better than the prediction effect of the logarithmic model. The results showed that the scheme of soil spectral reflectance prediction of soil conductivity was feasible, which provided an idea for the prediction of soil electrical conductivity by spectral information.

Abstract:In order to realize the dynamic monitoring of farmland ecosystem carbon flux, a method for estimating farmland ecosystem carbon flux based on Landsat series multi-source remote sensing data was proposed. Three experimental fields of agricultural Research and Development Center of University of Nebraska, northeastern United States were selected as the study area, and the corresponding flux site data published by AmeriFlux was used for subsequent modeling analysis. Based on the comprehensive analysis of climate variables, soil properties and plant traits, remote sensing factors closely related to carbon flux of farmland ecosystem were selected, and a full remote sensing factor data set covering key links of farmland ecological process was constructed. Then, the farmland carbon flux regression prediction model based on random forest was constructed. Compared with the ridge regression model and the lasso model, the model was more effective in estimating farmland ecosystem carbon flux, with a coefficient of determination of 0.94 and a root mean square error of 4.281g/（m2·d）. According to the importance analysis of factors based on random forest model, the contributions of DVI, NDWI, MSAVI, NRI and NDVI to carbon flux estimation were 35.6%, 25.8%, 12.2%, 7.8% and 5.2%, respectively. On the basis of above research, through the farmland ecosystem carbon balance space-time evolution characteristics analysis, the farmland carbon sink capacity was the strongest in 2013 when the crop growth was in the period of July and August in Nebraska, at the beginning of the planting soybeans and corn were rendered weak carbon source, and the carbon source ability was stronger for corn, in growth peak of corn and soybeans were in carbon sink, and the carbon sequestration ability was stronger for corn. The research result can provide theoretical support for accurately estimating the carbon budget of farmland ecosystems and guiding agricultural production.

Abstract:In order to improve the accuracy of cherry fruit recognition and the working efficiency of orchard automatic picking robot, totally 1816 sets of cherry original images collected in Yantai Academy of Agricultural Sciences and data images obtained with different data enhancement methods were used to establish the data set, the data set was divided into training set and test set according to rate of 8∶2, and YOLO v5 model was used to identify and detect cherry data sets enhanced by different data enhancement methods and combined enhancement methods based on the in-depth learning network. The results showed that offline enhancement and online enhancement had a certain positive effect on the improvement of model accuracy. The offline data enhancement strategy could significantly and stably increase the detection accuracy, and the online data enhancement strategy could slightly improve the detection accuracy. Using the combination of offline enhancement and online enhancement at the same time could greatly improve the average detection accuracy. In view of the mutual occlusion between cherry fruits and the difficulty in detecting small cherry targets in the picture, the detection accuracy of cherry fruits in the natural environment was low, the backbone network of YOLO v5 was changed, the transformer module with attention mechanism was added, and the neck structure was changed from the original pafpn to bifpn which could carry out two-way weighted fusion. The P2 module of shallow down sampling was added to the head structure. The experimental results showed that compared with other existing models and the improved YOLO v5 model with a single structure, the comprehensive improved model proposed had the highest detection accuracy, and the mAP@0.5∶0.95 was increased by 2.9 percentage points. The results showed that this method effectively enhanced the efficiency and accuracy of feature fusion in the recognition process, and significantly improved the detection effect of cherry fruit. At the same time, the trained network model was deployed on the Android platform. The system was simple and clear to use, and the requirements of user equipment environment were not high. Therefore, the system had certain practicability. It could detect cherry fruit in real time and accurately in the field environment, which laid a foundation for practical applications such as automatic service picking in the future.

Abstract:In order to reduce the influence of moisture and particle size on the soil organic matter prediction model established by the characteristic wavelengths selected in the traditional way, a method of extracting characteristic wavelengths was proposed. Sixty soil samples were collected from Shangzhuang Experimental Station of China Agricultural University, and the samples were naturally dried and divided into two, one portion was formulated into five particle size gradients (particle size of 2~2.5mm, 1.43~2mm, 1~1.43mm, 0.6~1mm, and 0~0.6mm), the other part was sieved through 0.6mm and formulated into five moisture gradients (5%, 10%, 15%, 20%, and 25% moisture content). The true values of soil organic matter content and soil spectral information were obtained by standard instruments, and the characteristic wavelengths were extracted by using the random frog-hopping algorithm. Totally seven wavelengths with high correlation with the true values of soil organic matter content were selected as the characteristic wavelengths under each moisture and particle size gradient, and multiple linear regression (MLR), partial least squares (PLS) and random forest (RF) models were established respectively. The results showed that the R2 of the modeling and prediction sets of the three models basically tended to decrease as the water content increased;the R2 of the modeling and prediction sets of the three models was the lowest in the gradient of 2~2.5mm, highest in the gradient of 0~0.6mm, and close to the R2 of the modeling and prediction sets in the rest of the gradient. Combined with the filter bandpass range of ±15nm, eight characteristic wavelengths of soil organic matter under moisture gradient were selected as the same or close to each other, and six characteristic wavelengths under particle size gradient were selected, and finally six wavelengths were eliminated under the 14 characteristic wavelengths determined under moisture gradient and particle size gradient by combining chemical bonding characteristics, and eight characteristic wavelengths were determined as follows: 932nm, 999nm, 1083nm, 1191nm, 1316nm, 1356nm, 1583nm, and 1626nm. The MLR, PLS and RF models were established respectively, and the results showed that the R2 of the modeling set and the R2 of the prediction set were not less than 0.8 and 0.75 for the three models established by the final selected organic matter characteristic wavelengths, and the best prediction effect was achieved by PLS, with the R2 of the modeling set and the R2 of the prediction set being 0.8809 and 0.8402, respectively. The model established had better applicability and prediction effect, and the influence of moisture and particle size on prediction was eliminated to a certain extent compared with the traditional way.

Abstract:Rapid acquisition of crop growth status information is essential for timely guidance of agricultural production. Based on the response mechanism of crop physiological and biochemical spectroscopy, a portable crop chlorophyll detection device based on ambient light correction was designed. The device measured reflectance spectral and ambient light spectral data with 20nm bandwidth centered at 610nm, 680nm, 730nm, 760nm, 810nm, and 860nm to calculate vegetation index and predict plant chlorophyll content. The features of the device were the supplemental light when the ambient light intensity was poor and the correction of the ambient light intensity under the supplemental light condition. To evaluate the sensor performance, the sensor was tested and calibrated. Experiments showed that the maximum difference in GPS positioning was 6.2m in latitude and 4.9m in longitude;the correlation between the light intensity response of the six bands of the spectral sensors and the measured values of the illuminance meter exceeded 0.99;the matching coefficients of the two spectral sensors were calibrated to 0.743 and 1.035 in the 610nm and 860nm bands, respectively. A fitting model between the supplemental light intensity and the measurement distance in the 610nm and 860nm bands was established for light environment correction;chlorophyll gradient experiments were conducted using nonwoven fabrics, and a mathematical model of NDVI vegetation index and plant chlorophyll content was established, with a model R2 of 0.685 under poor light environment conditions without supplemental light and a model R2 of 0.965 under supplemental light and with correction.

Abstract:In order to realize rapid, non-destructive and real-time monitoring of the leaf area index of cotton plants under different irrigation treatments, the canopy reflectance of cotton plants in four growth periods was obtained with the help of hyperspectral remote sensing technology, and the leaf area index of each cotton plant was obtained at the same time. The spectral preprocessing methods such as first-order derivation, second-order derivation, standard normal variate, multiple scattering correction and wavelet analysis were used to extract characteristic bands through continuous projection algorithm, PLS was used to establish hyperspectral estimation models for four growth periods and each growth period. Comparing the modeling accuracy of six pretreatment in four growth stages and each growth stage, it was shown that the wavelet decomposition scales of four growth stages, bud stage, flower stage and flower boll stage were 4, 2, 8 and 2, respectively, and the models were CWT-SPA-PLS, CWT-FD-SPA-PLS, CWT-SPA-PLS and CWT-FD-SPA-PLS respectively, which can achieve better accuracy;after SD treatment, better results were obtained in boll stage, R2 and RPD were 0.973 and 5.3295 respectively, which were better than other pretreatment results. The experimental results showed that the spectral information obtained by the preprocessing algorithm, especially the wavelet analysis method, can effectively estimate the leaf area index of cotton in four growth stages and each growth stage.

Abstract:A reasonable canopy structure and cultivation density of fruit trees can increase the amount of light interception within their canopies, which has an important impact on improving the yield and quality of fruit. A 3D point cloudbased model for predicting the canopy light distribution of group cherry trees was proposed by using a thin spindleshaped cherry tree as the research object. Firstly, the Azure Kinect DK camera was used to obtain the 3D point cloud data of the group cherry trees, and the complete 3D point cloud data of the group cherry trees were obtained through point cloud data preprocessing. Secondly, according to the actual cherry tree canopy segmentation method, the point cloud data of cherry tree canopies were point cloud stratified and the point cloud colour features of different regions were extracted. Again, a point cloud projection area calculation method based on the Delaunay triangulated concave packet algorithm was proposed to calculate the point cloud projection area of different regions through concave packet boundary point extraction and vector product fork multiplication. Finally, a model for predicting the light distribution in the canopy of group cherry trees was developed, which was a random forest model with point cloud colour characteristics and relative projected area characteristics as input and measured relative light intensity as output. The experimental results showed that the model was able to predict the light distribution in the canopy of cherry trees with a mean coefficient of determination of 0.885 and root mean square error of 0.0716. The research results can provide technical support for reasonable planting density management and automated pruning of cherry trees during dormancy.

Abstract:In order to study the prediction of crop water stress index (CWSI) of tomato canopy in greenhouse, through the deployment of multi parameter sensors, the environmental parameter inside and outside the greenhouse can be obtained in real time. Using gray correlation analysis, the correlation degree between environmental parameters and tomato canopy CWSI and the sub factor correlation coefficient between environmental parameters was calculated, the environmental parameters were sorted according to the correlation degree, and the impact on the accuracy of the model was considered. Finally, a total of seven parameters from nine environmental parameters were selected as the model input, and a prediction model of greenhouse tomato canopy crop water stress index (CWSI) based on LightGBM was established. Combined with Bayesian algorithm to optimize the key parameters, the correlation between the prediction results of the model and the CWSI value calculated by Jones empirical formula was analyzed. Under the same computing environment, it was compared with GBRT and SVR models respectively. The experimental results showed that the coefficient of determination (R2), mean absolute error (MAE), root mean square error (RMSE) and operation time of the Bayesian optimized LightGBM model were 0.9601, 0.0218, 0.0314 and 0.0518s, respectively. Compared with GBRT and SVR models, R2 was increased by 2.14% and 14.05% respectively, MAE was reduced by 0.0093 and 0.0612 respectively, RMSE was reduced by 0.0097 and 0.0591 respectively, and the time was shortened by 0.0459s and 0.0612s respectively. It was showed that the LightGBM model proposed had better performance, which could effectively improve the prediction accuracy of greenhouse tomato canopy CWSI, and provide a strategy for realizing greenhouse tomato on-demand irrigation and a reference for water requirement research.

Abstract:Ruminants contribute to global warming by releasing methane gas into the atmosphere. Methane, the second largest greenhouse gas in the world, has a greenhouse effect of 25 times of that of CO2. The accounting methods and monitoring technologies of ruminant methane emissions at home and from the overseas are developing in the direction of comprehensiveness, intelligence and precise management and control. In order to monitor the methane emissions of ruminants and provide a reference for the accounting of methane emission reduction in the livestock and poultry breeding industry, it has become a hot research topic for relevant scholars to find suitable accounting and monitoring methods for methane emissions from ruminants. The sources of ruminant greenhouse gas methane emissions, accounting methods, and the application status of monitoring technologies were mainly reviewed. Referring to relevant domestic and foreign literatures on methane emissions from livestock and poultry breeding, the advantages and disadvantages of the methods of Organization for Economic Co-operation and Development (OECD), United Nations Intergovernmental Panel on Climate Change (IPCC) coefficient method, provincial greenhouse gas inventory compilation guidelines and life cycle assessment (LCA) were compared. The research progress of ruminant methane emission monitoring technology application of seven methods, including respiratory mask and breathing head box monitoring method, infrared spectroscopy monitoring method, and respiratory metabolic chamber monitoring method, was compared and analyzed, in order to promote the domestic ruminant greenhouse gas methane monitoring technology. The research result can provide reference for development and emission reduction work.

Abstract:Biomass stoves are environmentally friendly stoves that provide heat through biomass combustion, which have a long history and extensive foundation in China. It has the advantages of high efficiency, environmental protection, high thermal efficiency, low smoke and dust emission, etc. It is widely used in domestic hot water, building heating and industrial heating and other fields. In the combustion process, however, due to technical limitations, it is difficult to ensure the full combustion of biomass fuels, and problems such as low boiler thermal efficiency and environmental pollution have become constraints on the development of biomass stoves in China. To solve this problem, a biomass molding combustion stove was proposed based on variable pressure jet combustion technology. During the combustion process, the primary air and secondary air formed by the two air intake mechanisms of the stove were in the furnace body in the form of spin, integral rotation or superposition of the two, changing the air volume ratio and air intake position of the secondary air and the inflation angle. As a result, different air pressure jets were formed in the stove to strengthen the heat and mass transfer, ensure the burnout effect, and effectively control the generation and emission of nitrogen oxides during the combustion process. The fuel combustion test results showed that the combustion thermal efficiency of corn stover was 85.97%, the bottom ash slagging rate was 3.42%, and the biomass particles were fully burned;the SO2 emission was 18.34mg/m3, the NOx emission measurement was 91.45mg/m3, and the CO volume fraction was 0.092%. The pollutant discharge met the national standard. The boiler had good stability and met the design requirements. 

Abstract:As the grape production increases year by year, the quality detection of grapes in the field becomes more and more important to improve the economic benefits after flowing into the market. The traditional method of external quality detection, which mainly relies on the observation of workman, introduces non-negligible errors. The intrinsic quality detection is considered as destructive and inefficient by using the method of sugar level testing of grapes. With the development of deep learning and image processing technology, the field grape quality detection based on machine vision overcomes the limitations of traditional manual inspection and has the advantages of fast, accurate, realtime and lossless. According to grape varieties and quality evaluation indicators, a systematical analysis and summary of the research related to the nondestructive quality detection method of grapes in the field was provided based on machine vision technology. The main body consisted of two parts, which were machine vision detection methods of grape varieties and machine vision detection methods of grape quality. The common factors affecting the quality of grapes were obtained on the basis of the analysis of different grape variety evaluation factors. The intrinsic quality factors included soluble solids, total acid, total phenol and moisture content while the external quality factors included fruit size, quantity, color, and disease defects and so on. Several methods of grape variety identification based on fruit and leaf were introduced, including canonical correlation analysis, support vector machine, and deep learning. The detection method based on fruit characteristics was more accurate, while the detection method based on leaf characteristics can be applied to a longer growth period. As the variety of grapes differred, the standard of their internal and external quality also varied. A detailed summary of the research related to the non-destructive quality detection methods for the intrinsic quality and external quality of grapes in the field was provided. For the quality detection of grapes, the comparison was conducted between the traditional morphological methods such as thresholding, the edge contour search and the corner detection algorithm with the deep learning methods such as Mask R-CNN. It was concluded that the deep learning detection method held the advantages of strong scalability, fast detection speed and high accuracy. In addition, the application principle and advantages and disadvantages of nearinfrared spectroscopy and hyperspectral imaging technology in intrinsic quality detection were summarized. Hyperspectral technology outperformed in terms of accuracy, while nearinfrared spectroscopy technology had lower cost and faster analysis speed. In the field of non-destructive quality detection of grapes, machine vision algorithms based on spectral analysis still faced the challenges of complex field grape growth environment and variable daytime light. Finally, in view of the difficulty of image acquisition, insufficient multidimensional image information, and weak foundation of detection instruments faced by nondestructive quality detection methods of grapes in the field, it was proposed that it was necessary to improve the intelligent equipment for data collection and analysis while improving the machine vision algorithm, thus providing efficient tools combining software and hardware for the quality detection of grapes in the field.

Abstract:In view of the demand for online detection of sprouted potatoes, a lightweight convolutional neural network was proposed to detect sprouted potatoes. Firstly, the acquired potato samples were collected based on the grading line, and the samples were expanded through data enhancement. The Shuffle-Net lightweight convolutional neural network was built, and the effects of different learning rates and learning rate decay strategies on the model were compared. Experiment results showed that when the learning rate was 0.001 and the decay strategy was W-EP, the performance was the best. The overall recognition accuracy of sprouted potato and healthy potato was 97.8%, the single sample recognition time was 0.14s, and the model memory footprint was 5.2MB. The experimental results were evaluated, the precision was 98.0%, the recall was 97.1%, the specificity was 98.4%, and the harmonic mean was 97.5%. The VGG11, Alex-Net, and Res-Net101 models were selected for comparison with the model. It was found that the recognition accuracy of the model was greatly improved compared with that of the VGG11 and Alex-Net, and the recognition speed of a single sample was 5 times higher than that of Res-Net101. Compared with VGG11, it was nearly 7 times higher, and the model volume was greatly reduced compared with that of VGG11, Alex-Net, and Res-Net101. In the experiment, the internal convolution of the model was visually analyzed and the results were misjudged. It was found that when the buds were dark, short and at the edge of the tuber, misjudgment would be caused. It can be concluded that this experimental model realized the accurate and effective identification of sprouted potato, and it also had the advantages of fast identification speed, small size and strong portability, which can provide theoretical support for the external nondestructive testing and classification of agricultural products.

Abstract:Sugar content is one of the important indicators for watermelon grading, for the drawbacks of traditional watermelon detection methods, the feasibility of acoustic characteristics combined with machine learning for non-destructive detection and grading of watermelon was investigated. The acoustic detection system of watermelon was designed and the time domain signals of different batches of samples were collected. After the time domain signal was normalized, the frequency domain signal was obtained by fast Fourier transform and pre-processed by detrending. The principal components of the frequency domain signal were extracted by using principal component analysis, the cumulative contribution rate of the first three principal components was 95.32%, the samples with different levels were differentiable using the first and second principal components. Watermelon all-variable grading models were developed by using four different machine learning algorithms, and the prediction set classification accuracies all reached over 66%. Feature variables were extracted by using stability competitive adapative reweighted sampling algorithm, which reduced the number of variables by about 84%. The performance of the classification models developed using the extracted feature variables were all improved, with the support vector machine model achieved the highest prediction set accuracy (95.56%), F1 score (96%) and Kappa coefficient (93%). The results indicated that acoustic characterization combined with machine learning was feasible for non-destructive detection and grading of watermelons. The research result can provide a feasible technical solution for non-destructive detection and grading of watermelon, and provide a reference for non-destructive detection and grading of other similar fruits and vegetables.

Abstract:Aiming at the operation automation and high-precision safety operation requirements of trailer-type large-load special vehicles, a set of trailer-type unit autonomous navigation control system was designed for the large-load tractor. The vehicle-mounted program control system adopted a modular distributed system, and the communication between the modules in the system was realized through the CAN bus. The TCP protocol was used for data communication between the remote operation and management platform and the vehicle-mounted program control terminal, so as to realize the information exchange between the remote operation and management platform and the vehicle-mounted program control system. The kinematic model of the trailer-type special vehicle was established, and the towing state and the minimum turning radius of the towed machine were analyzed during the towing operation. Aiming at the defect of fixed foresight distance in the traditional pure pursuit algorithm, the foresight distance was dynamically calculated based on the current realtime speed of the tractor, tracking path curvature, heading and other information, and the fixed foresight distance into dynamic foresight distance tracking was improved. The control parameters were optimized for multiple parameters, which significantly improved the accuracy of trajectory tracking;the random forest algorithm was used to extract the features of the trajectory tracking data, and the algorithm parameters were modified according to the weight of each feature importance index. On an open concrete runway with a maximum horizontal inclination of 2° on the test site, the mass of the tractor was 2t, the mass of the towed machine was 10t, and the length was 22m, and the design of the traction unit required a maximum running speed of 6km/h and a maximum lateral deviation of 50cm. According to the long and complex path test and data analysis, the maximum delay of the system correction response was 84ms, the maximum absolute lateral error of the traction unit was 37.14cm, and the average absolute error was 14.91cm, which met the practical application requirements in large-load towing operations.

Abstract:In order to study the influence of tractor rotation tillage load on power take-off driveline, the system dynamics modeling, bench experiment verification, field test and simulation analysis were used. Firstly, based on the structure analysis of the power output driveline, a torsional vibration coupled spatial dynamics model was established to describe the load transfer mechanism. In this model, transverse and vertical transmission effects of gear meshing were considered in detail. Secondly, the tractor power take-off (PTO) loading experiment bench was used to verify the simulation results of the model. The verification results showed that the maximum lateral and vertical mesh frequency errors were 4.24% and 5.12%, respectively, which met the modeling requirements. Then, the data acquisition system composed of wireless torque sensor and BeiDou positioning system was built, and the data of rotary tillage operation in the field under L1 (2.07km/h), L2 (3.10km/h) and L3 (5.29km/h) were collected respectively. The results of the field test showed that the load level and fluctuation range of rotary tillage were increased with the increase of gear and driving speed. Finally, the influence of PTO load on gear transfer characteristics was simulated by using the established dynamic model. The results showed that the higher the tractor gear for rotary tillage operation, the greater the vibration displacement of power takeoff driveline caused by the fluctuation of PTO load, which was mainly reflected in the lateral vibration. The above research provided theoretical model and data reference for durability design and load spectrum loading experiment of tractor power takeoff driveline.

Abstract:The humanized and less humanized operation is still the main working mode of the current mechanized operation. As the core common component of the highend agricultural machinery, the integrated intelligent console is widely used in tractors, sprayers, harvesters and other large agricultural machinery equipment. As a medium of direct contact with the driver, it has a direct impact on the physical and mental health of the driver. Aiming at the problem of high strength and high psychological load of manual operation, and the requirements of operators’ active health, the domestic typical console was selected. Based on Kansei engineering theory,the psychological load index system was constructed, and the static manmachine semiphysical test of tractor intelligent console was carried out. The principal component model of psychological load was established, and the console based on minimum psychological load was optimized. Firstly, the psychological load evaluation system was constructed, and the manmachine semiphysical test bed was built based on the domestic large tractor. The actual driving video of ploughing condition was selected and the test scheme was designed. Secondly, totally 10 drivers with tractor driving experience were selected to test and record the psychological load index data. Thirdly, the principal component model of psychological load was established according to the experimental data, and the design defects of the psychological load of the console before optimization were analyzed. Finally, based on the results of principal component analysis, in order to minimize the psychological load, the manmachine engineering theory was applied to optimize the console from the aspects of interface element distribution, element color and shape. The optimized console was obtained and then verified by experiments.The results showed that the psychological load of the driver was mainly composed of fatigue perception factors and fatigue mitigation factors. The average psychological comfort score of the console before optimization was 0.403. The main fatigue perception factor was visual and neck fatigue (weight: 0.458), and the main fatigue mitigation factor was job selfefficacy (weight: 0.578). The average score of the optimized console psychological comfort was increased to 2.048, which can significantly alleviate the psychological load. The research results can provide some reference for the design of comfort of intelligent tractors, and it can help to improve the design theories of agricultural machinery equipment active health.

Abstract:As an important tool for agricultural energy conservation and emission reduction, electric tractors are also one of the development and transformation directions of agricultural machinery equipment products in the future. A lot of technical research and product trials have been carried out at home and abroad. Starting with the concept of electric tractor technology, the stages of electric tractor technology development and its typical products are sorted out, and the technical characteristics of each stage of development and the main factors affecting the development of electric tractor technology are summarized. The development of battery technology directly affected the power supply mode and watt level as well as the change of the energy system model of electric tractors. The energy density and the cost of use were the main factors that affected the tractor’s pure electrification. Compared with automobiles, tractors had higher load rates during plowing and rotary tillage. At present, the energy density level that can be achieved by battery products had a certain gap in supporting electric tractors in heavy farmland operations effectively and independently. The development of lowpower pure electric dedicated plant protection models or the use of multiple energy systems on largerpower tractors was in line with the current state of battery technology and can provide common technical reserves for the realization of pure electric tractors in the future. Based on the existing research results at home and abroad, the current research status of the scheme and design, control technology, simulation and test technology of electric tractor research is analyzed, and the characteristics of existing research results in various directions are summarized. It is necessary to continue to form and improve the theoretical system of electric tractors in the future, so as to gradually achieve universal and standardized electric tractor whole machine and implements, agronomy, energy system, new energy generation and rural energy storage technology and application technology matching and collaborative control. Combined with technologies such as Internet of Things, intelligent control and automatic driving to improve the intelligence level of electric tractors, the research results can provide reference for the development of related technologies.